“Anthropic is investigating the security breach of its Mythos model but has not yet found evidence of malicious use.”